Single or combined immune checkpoint inhibitors compared to first‐line chemotherapy with or without bevacizumab for people with advanced non‐small cell lung cancer
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
Primary
To determine the efficacy and safety of first‐line immune checkpoint inhibitors, or immuno‐oncology therapy (IO), as monotherapy or in combination for patients with advanced non‐small cell lung cancer (NSCLC).
Secondary
To maintain the currency of evidence by taking a living systematic review approach.
Background
Description of the condition
Lung cancer is one of the leading causes of cancer death worldwide (Bray 2018). Non‐small cell lung cancer (NSCLC), which accounts for 85% to 90% of lung cancers (Novello 2016), includes two major histological types: non‐squamous carcinoma and squamous carcinoma. Among Caucasians, about 20% of NSCLCs harbour a potential targetable driver (Barlesi 2016). A total of 23% of NSCLCs express programmed death‐1 ligand (PD‐L1) with a tumour proportional score (TPS) ≥ 50%, 38% with a TPS of 1% to 49%, 50% with a TPS ≥ 5%, and 39% with a TPS < 1% (Garon 2015;Gettinger 2016).
Until recently, platinum‐based chemotherapy with or without bevacizumab, an antiangiogenic agent, represented the standard first‐line treatment in non‐oncogene‐addicted NSCLC, achieving median progression‐free survival (PFS) of six to eight months, and median overall survival (OS) of 12 months (Gridelli 2014; Perez‐Moreno 2012).
Description of the intervention
The arrival of immune checkpoint inhibitors (IO), has dramatically changed the treatment paradigm in the first‐line setting.
In 2016, pembrolizumab, a programmed cell death protein‐1 (PD‐1) inhibitor, was approved by the Food and Drug Administration (FDA) and the European Medicines Agency (EMA) as monotherapy in treatment‐naive metastatic NSCLCs with a PD‐L1 TPS ≥ 50%. Approval was granted on the basis of a phase III trial comparing pembrolizumab to platinum‐based chemotherapy (KEYNOTE 024), which showed improvement in PFS, overall response rate (ORR), and health‐related quality of life (HRQOL) in favour of pembrolizumab, and a significant OS advantage despite a high cross‐over rate (62%) (Reck 2016).
Recently, first‐line single‐agent pembrolizumab has been shown to significantly improve OS compared to standard platinum‐based chemotherapy in three pre‐specified patient groups according to tumour PD‐L1 expression score: ≥ 50%, ≥ 20%, and ≥ 1%. An exploratory analysis showed no significant difference in OS (hazard ratio (HR) 0.92, 95% confidence interval (CI) 0.77 to 1.11) among patients with PD‐L1 expression of 1% to 49%. In addition, results show no significant difference in PFS in any of the subgroups (Lopes 2018).
Nivolumab, another PD‐1 inhibitor, failed to show a PFS advantage in a PD‐L1 selected population (TPS ≥ 5%) compared to first‐line standard chemotherapy. Post hoc analysis revealed no difference in PFS among patients with PD‐L1 TPS ≥ 50%. An additional exploratory analysis showed improvement in ORR and PFS among patients with high tumour mutational burden (TMB), defined as the presence of 243 or more somatic missense mutations in tumour samples. However, no OS benefit was noted in the TMB selected population, perhaps because of the high cross‐over rate in the chemotherapy arm (68%) (Carbone 2017).
The combination of PD‐L1 inhibitors and anti‐cytotoxic T‐lymphocyte‐associated protein 4 (CTLA4) agents has also been investigated. A phase III multi‐arm study showed that the combination of nivolumab plus ipilimumab compared to standard chemotherapy significantly improved ORR and PFS in NSCLC harbouring high TMB (≥ 10 mutations per megabase) regardless of PD‐L1 expression (Hellmann 2018). More recently, a press release reported similar OS results for patients with high TMB and low TMB treated with the nivolumab and ipilimumab combination (www.bms.com).
A trial of tremelimumab (another CTLA4 inhibitor) in combination with durvalumab, an PD‐L1 inhibitor, did not show significant improvement in PFS compared to standard chemotherapy in advanced NSCLC (www.astrazeneca.com).
Regarding safety, single‐agent anti‐PD‐1/PD‐L1 agents have demonstrated a manageable toxicity profile, with grade 3 to 5 adverse events ranging from 9.5% to 17.8% (Carbone 2017;Lopes 2018). However, this rate may rise to ˜ 30% in people with NSCLC treated with the association of ipilimumab and nivolumab (Hellmann 2018).
Special populations, such as people with uncontrolled brain metastases, autoimmune disorders, steroid dependency, and poor performance status, usually are not included in randomised clinical trials; furthermore, people with oncogene‐addicted (i.e. epidermal growth factor receptor (EGFR) mutated or anaplastic lymphoma kinase (ALK) rearranged) NSCLC have been excluded from randomised trials testing single‐agent or combined immune checkpoint inhibitors. Therefore, this review will not attempt to address specific questions for these subgroups of patients. However, particular attention will be paid to potentially interesting clinical and pathological variables that may influence the outcomes of immune checkpoint inhibitors, such as sex, age, smoking status, and PD‐L1 expression. In fact, subgroup analyses from randomised clinical trials in people with pre‐treated NSCLC have raised doubts about the immunotherapy benefit for elderly people (i.e. those older than 75 years) (Brahmer 2015), or among those with NSCLC with low/negative PD‐L1 expression (Borghaei 2015). Furthermore, a recent meta‐analysis showed that in patients with cancer, the magnitude of benefit may be sex dependent, with worse outcomes reported for females (Conforti 2018).
How the intervention might work
Checkpoint inhibitors comprise a class of humanised immunoglobulins that target and inhibit molecules responsible for the physiological 'off‐switch' of immune cells to prevent an excessive and uncontrolled immune response. Their inhibition activates T cells and enhances the adaptive anti‐cancer immune response.
Both nivolumab and pembrolizumab (immunoglobulin (Ig)G4 monoclonal antibodies) bind PD‐1 on immune cells, blocking their interaction with PD‐L1 and PD‐L2 expressed by tumour cells (Ishida 1992). Atezolizumab, durvalumab, and avelumab (IgG1 monoclonal antibodies) bind PD‐L1 on tumour cells, preventing interaction with PD‐1. Both classes of drugs activate PD‐1 pathway‐mediated inhibition of the immune response.
CTLA4 is expressed by T cells and, after binding to CD80/CD86, activates an inhibitory downstream signal in human lymphocytes (Hathcock 1993). Ipilimumab (IgG1) and tremelimumab (IgG2) block human CTLA4, inducing T‐cell activation, proliferation, and intratumoural infiltration, with improved anti‐cancer immune response.
Combining PD‐L1 inhibitors and anti‐CTLA4 agents might improve antitumour immunity because PD‐1 and CTLA4 modulate effector T‐cell activation, proliferation, and function through distinct complementary mechanisms (Okazaki 2013). Double immune checkpoint blockade may have a relevant role for all high TMB tumours (Lawrence 2013), which are known to be highly sensitive to immunotherapy. Furthermore, recent evidence has shown that a primary target of PD‐1 inhibition is CD28, a co‐stimulatory receptor that can bind to CD80/CD86 (Hui 2017). So the combination of CTLA4 and PD‐1 inhibitors may have a synergistic effect with high activation of CD28/CD80 or the CD86 axis and an increased antitumour immune response.
Why it is important to do this review
Recent advances in immunotherapy have led to approval of immunotherapy alone or in combination with chemotherapy as first‐line treatment for NSCLC, according to PD‐L1 expression. Double immune checkpoint blockade is also emerging as a treatment option in NSCLC with high TMB. Some questions remain unanswered, such as the best treatment strategy (monotherapy or combination therapy) and the respective role of different biomarkers (PD‐L1, TPS, TMB) for patient’ selection.
Living systematic review approach
After publication of this review, we are planning to maintain it as a living systematic review (LSR). This means that we will search the literature continually and will incorporate any new evidence as it becomes available (Elliott 2017). A living systematic review approach is appropriate for this review because the topic is important for clinical decision‐making; current evidence is unlikely to provide certainty; and a very active research programme is ongoing. Indeed, the review authors are aware of several ongoing trials and believe that incorporating results from these trials in a timely manner is crucial, as these findings may have an impact on review conclusions.
Objectives
Primary
To determine the efficacy and safety of first‐line immune checkpoint inhibitors, or immuno‐oncology therapy (IO), as monotherapy or in combination for patients with advanced non‐small cell lung cancer (NSCLC).
Secondary
To maintain the currency of evidence by taking a living systematic review approach.
Methods
Criteria for considering studies for this review
Types of studies
We will include randomised controlled trials (RCTs) reporting on the efficacy or safety of immune checkpoint inhibitors, or immuno‐oncology therapy (IO), as first‐line treatment for people with advanced NSCLC with or without blinding. We will apply no language or publication status restrictions, and if sufficient data are available, we will consider including meeting abstracts and unpublished online data.
If we consider inclusion of cross‐over studies, we will include them only at phase I to avoid a 'carry‐over effect' from the first intervention to the second phase.
Types of participants
We will include studies involving participants with metastatic NSCLC or locally advanced NSCLC not susceptible to curative treatment. Patients should have not received any first‐line systemic treatment. We will not apply any restriction with regard to age, gender, drug dosage, or treatment duration.
Types of interventions
We will consider studies for inclusion if researchers report one or more the of the following comparisons.
-
Single‐agent immune checkpoint inhibitor (IO) versus standard first‐line therapy (doublet chemotherapy ± bevacizumab).
-
Doublet immune checkpoint inhibitors (IO) versus standard first‐line therapy (doublet chemotherapy ± bevacizumab).
-
Doublet immune checkpoint inhibitors (IO) versus single‐agent immune checkpoint inhibitor in a first‐line setting.
A doublet chemotherapy regimen includes any platinum‐based doublet along with a third‐generation agent (i.e. gemcitabine, vinorelbine, taxanes, pemetrexed).
Although we acknowledge that a lot of evidence is available on the combination of first‐line immune checkpoint inhibitors and chemotherapy, an ongoing Cochrane systematic review and meta‐analysis will examine the potential benefit of immunotherapy and chemotherapy combinations versus first‐line chemotherapy or single‐agent IO (Syn 2018).
Types of outcome measures
Primary outcomes
-
Overall survival (OS): defined as time from randomisation to death from any cause
-
Progression‐free survival (PFS): defined as time from randomisation to cancer recurrence or death from any cause
Secondary outcomes
-
Overall objective response rate (ORR): measured by Response Evaluation Criteria in Solid Tumours (RECIST) v.1.1 (Eisenhauer 2009); guidelines for response criteria for use in trials testing immunotherapeutics (iRECIST) (Seymour 2017); or immune‐related RECIST (irRECIST) (Nishino 2013)
-
Health‐related quality of life (HRQOL): measured via validated generic or disease‐specific questionnaires, or validated items
-
Treatment‐related adverse events (AEs): any AEs as reported by the included trials individually. We will investigate the incidence of grade 3 (severe or medically significant but not immediately life‐threatening; hospitalisation or prolongation of hospitalisation indicated; disabling; limiting self‐care activities of daily living) and grade 4 events (life‐threatening consequences; urgent intervention indicated) based on the Common Terminology Criteria for Adverse Events (CTCAE) and Patient‐Reported Outcomes CTCAE (PRO‐CTCAE) (Kluetz 2016). We will also check the included trials for incidence of grade 5 AEs and death related to adverse events
Search methods for identification of studies
Electronic searches
We will search the following electronic databases from inception.
-
Cochrane Lung Cancer Group Trial Register.
-
Cochrane Central Register of Controlled Trials (CENTRAL), in the Cochrane Library.
-
MEDLINE, accessed via PubMed.
-
Embase.
We will apply no restriction on language of publication.
We have presented the search strategies for CENTRAL, MEDLINE, and Embase in Appendix 1, Appendix 2, and Appendix 3, respectively.
We will search all databases using both controlled vocabulary (namely, medical subject headings (MeSH) in MEDLINE and EMTREE in Embase) and a wide range of free‐text terms. We will perform the MEDLINE search using the Cochrane highly sensitive search strategy and precision‐maximising version (2008 version), as described in the Cochrane Handbook for Systematic Reviews of Interventions (Chapter 6.4.11.1, and detailed in Box 6.4.b) (Higgins 2011b).
We will also conduct searches in the following clinical trials registries to identify unpublished and ongoing trials.
Living systematic review approach
In approaching this as a living systematic review, we will search the following databases monthly, using auto‐alerts when possible.
-
Cochrane Lung Cancer Group Trials Register.
-
Cochrane Central Register of Controlled Trials (CENTRAL), in the Cochrane Library.
-
MEDLINE, accessed via PubMed.
-
Embase.
For other electronic databases and electronic sources (including trials registries), we will set up auto‐alerts (when possible) to deliver a monthly search yield by email. When auto‐alerts are not possible, we will manually search these electronic databases and electronic sources each month.
We will review search methods and strategies at least yearly, to ensure that they reflect any terminology changes in the topic area or in the databases. For example, we will review search methods and strategies whenever researchers become aware of a new drug meeting the definition for either intervention or comparison and used for the review population.
Searching other resources
We will handsearch the references of eligible studies to identify additional studies for inclusion.
We will search the meeting abstracts of conferences at the following sources from 2015 onwards.
-
World Conference on Lung Cancer (WCLC).
-
European Society for Medical Oncology (ESMO).
-
European Society for Medical Oncology Immuno‐Oncology congress (ESMO IO).
-
European Lung Cancer Conference (ELCC).
-
American Society of Clinical Oncology (ASCO).
-
American Association of Cancer Research (AACR).
We will also retrieve clinical study reports about the checkpoint inhibitors from the European Medicines Agency (EMA) website.
Living systematic review considerations
In developing this living systematic review, we will note when these key conferences are to be held and will search conference proceedings when published.
We will contact corresponding authors of ongoing studies as we identify them and will ask them to advise when study results are available, or to share early or unpublished data. We will contact the corresponding authors of any newly included studies for advice regarding other relevant studies.
We will manually search the reference lists of all newly included studies.
Data collection and analysis
Selection of studies
Two review authors will screen independently all titles and abstracts retrieved by electronic searches. These review authors will obtain the full texts for all relevant studies and will check independently the eligibility of each study against review eligibility criteria. We will pursue discordant evaluations by discussion to reach consensus. When necessary to reach consensus, we will involve a third review author.
We will immediately screen all new citations retrieved by our searches.
Data extraction and management
The review authors will develop a data extraction form using Covidence. Two review authors (RM, RF) will independently extract relevant data. To reach consensus, we will involve a third review author when necessary (MI). We will not be blinded to the names of study authors nor to the institutions where studies were conducted and funded. When we encounter multiple publications for the same study, we will choose the first publication dealing with the primary endpoint in this review as a study identifier (study ID).
We will extract the following details from each included study.
-
Source: citation, study name if applicable, and contact details.
-
Study details: study design, location, setting (type and stage of disease), sample size, and study period: dates of first and last included participants, date of last follow‐up.
-
Characteristics of participants: inclusion and exclusion criteria, number of participating centres, number of participants, and participant and tumour characteristics (age, sex, ethnicity, smoking status, performance status, histology, molecular status, tumour‐node‐metastasis (TNM) stage, PD‐L1 expression, TMB).
-
Characteristics of interventions (e.g. drugs, doses, cycle duration).
-
Outcomes: primary and secondary outcomes with definitions and time points.
-
Results: number of participants allocated to each group, and for each outcome of interest, sample size, missing participants, dropout rate, summary data for each group, estimate of effect with confidence interval and P value and subgroup analyses, and whether analyses have been performed by intention‐to‐treat (ITT) or per‐protocol methods.
-
Miscellaneous: funding source.
Assessment of risk of bias in included studies
Two review authors (RM, RF) will independently apply the Cochrane 'Risk of bias' tool per Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions, to assess quality and potential biases across studies eligible for inclusion in this review (Higgins 2011). We will rate each domain of the tool as having 'low', 'high', or 'unclear' risk of bias at study level and for each outcome if possible, and we will support the rating of each domain by providing a brief description. We will summarise risk of bias for each outcome within a study by considering all domains relevant to the outcome (i.e. both study‐level entries, such as allocation sequence concealment, and outcome‐specific entries, such as blinding). We will provide a figure to summarise the risk of bias, similar to Figure 8.6.C, as presented in the Cochrane Handbook for Systematic Reviews of Interventions.
When two review authors cannot reach consensus, we will consult with a third review author (SPB).
Using the Cochrane 'Risk of bias' tool, we will consider the following domains.
-
Selection bias: random sequence generation.
-
Selection bias: allocation concealment.
-
Performance bias: blinding of participants and personnel.
-
Detection bias: blinding of outcome assessment.
-
Attrition bias: incomplete outcome data for outcomes related to efficacy and safety.
-
Reporting bias: selective reporting of outcomes.
-
Other bias, such as inclusion of patients concordant to pre‐specified number of participants needed for calculation, unplanned interim analyses, and unbalanced baseline characteristics across arms.
Measures of treatment effect
For time‐to‐event outcomes ‐ OS and PFS ‐ we will use hazard ratios (HRs) to measure treatment effects. We will report each HR along with the 95% confidence Interval (CI). An HR of one indicates that the hazard rate is equivalent between experimental and control groups, and an HR other than one indicates differences in hazard rates between the two groups. We will extract the HR from the included studies when it is available. When it is not reported in the included study, we will try to calculate the HR by using Kaplan‐Meier survival curves and the dedicated methods of Parmar and Tierney (Parmar 1998;Tierney 2007).
For dichotomous outcomes ‐ AE and ORR ‐ we will use risk ratios (RRs) and 95% CIs if possible.
For dichotomous outcomes related to OS and PFS at specific time points, we will use survival rates and 95% CIs.
For continuous outcomes (HRQOL), we will use mean differences (MDs) between treatment arms when a similar scale was implemented to measure outcomes, and standardised mean differences (SMDs) when different scales were used to measure the same outcome. We will confirm that higher scores for continuous outcomes have the same meaning for the particular outcome, will explain the direction, and will report if directions were reversed.
Unit of analysis issues
The primary unit of analysis will be the participant.
Studies with multiple treatment groups
For studies with multiple comparison groups that compare two or more intervention groups versus the same control group, we will first try to combine groups to create a single pair‐wise comparison. We will calculate within‐study correlation as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).
When studies employ a cross‐over design and provide sufficient reporting, we will follow the recommendations detailed in Chapter 16.4.5 in the Cochrane Handbook for Systematic Reviews of Interventions (Elbourne 2002; Higgins 2011).
Dealing with missing data
When we identify missing or unclear data, we will contact the study author directly. We will follow Cochrane recommendations when dealing with such data details, as provided in Chapter 16 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), and we will consider two approaches.
-
Analysing only available data.
-
Imputing the missing data using replacement values and treating these as if they were observed.
Assessment of heterogeneity
We will follow Cochrane recommendations for assessment of heterogeneity (Deeks 2011). We will visually investigate heterogeneity by using forest plots generated via RevMan 5.3 (RevMan 2014). We will assess statistical heterogeneity of treatment effects between pooled trials for each considered outcome by using the I² statistic to quantify the degree of heterogeneity (Higgins 2002), and we will consider I² > 30% as showing moderate heterogeneity, with I² > 75% signifying significant heterogeneity. When at least moderate heterogeneity is underlined, we will perform and will report the results of both fixed‐effect and random‐effects meta‐analyses, and we will explore heterogeneity by performing pre‐specified subgroup analyses.
Assessment of reporting biases
We are planning to generate funnel plots and to perform Egger's linear regression tests to investigate reporting biases for considered outcomes when the number of trials included in a single meta‐analysis is sufficient (at least 10 trials). We will follow recommendations provided in Chapter 10 of the Cochrane Handbook for Systematic Reviews of Interventions (Sterne 2011).
Data synthesis
If sufficient clinically similar studies are available, we will pool their results in meta‐analyses and will perform meta‐analyses based on ITT analyses when available.
We will perform meta‐analyses according to recommendations given in Chapter 9 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). For meta‐analyses, we will enter data into Revman 5.3 (RevMan 2014). A review author (RM) will enter the data, and a second review author (RF) will double‐check the data for accuracy.
We will consider the results obtained from fixed‐effect and random‐effects models. We will base our choice between the two models on assessment of heterogeneity.
We will apply the inverse‐variance method for fixed‐effect models for time‐to‐event outcomes. We will apply the Mantel‐Haenszel method for dichotomous outcomes and the inverse‐variance method for continuous outcomes. We will use Peto’s odds ratio (OR) method under the fixed‐effect model in cases of rare events (Brockhaus 2014). For random‐effect models, we will apply the DerSimonian and Laird method (DerSimonian 1986).
We will follow the GRADE approach when creating a ’Summary of findings’ table, as suggested in Chapters 11 and 12 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).
We will use the five GRADE considerations to rate the quality of evidence as ’high’, ’moderate’, ’low’, or ’very low’.
-
Risk of bias: serious or very serious.
-
Inconsistency: serious or very serious.
-
Indirectness: serious or very serious.
-
Imprecision: serious or very serious.
-
Publication bias: likely or very likely.
We will include the following outcomes in the ’Summary of findings’ table.
-
OS.
-
PFS.
-
ORR.
-
HRQOL
-
AEs: grades 3, 4, and 5.
Living systematic review considerations
Whenever we identify new evidence (meaning new studies, data, or information) that is relevant to the review, we will extract the data and assess risk of bias, as appropriate. We will not adjust meta‐analyses to account for multiple testing, given that methods related to frequent updating of meta‐analyses are under development (Simmonds 2017).
We will wait until accumulating evidence changes one or more of the following components of the review before incorporating it and re‐publishing the review.
-
Findings of one or more outcomes (e.g. clinically important change in size or direction of effect).
-
Credibility (e.g. GRADE rating) of one or more outcomes.
-
New settings, populations, interventions, comparisons, or outcomes studied.
Subgroup analysis and investigation of heterogeneity
We will perform subgroup analyses, if data are adequate, to assess the effect on heterogeneity for each of the primary and secondary outcomes. We are planning the following subgroups.
-
Immune checkpoint inhibitor or immuno‐oncology therapy (IO) class (anti‐PD‐1, anti‐PD‐L1, anti‐CTLA4).
-
Inclusion or exclusion of participants with brain metastasis.
-
Follow‐up duration.
-
Tumour PD‐L1 expression score ≥ 50%, ≥ 20%, or ≥ 1%, or TMB expression (high vs low).
-
Participant‐related prognostic factors such as age, sex, performance status, smoking status, and liver metastasis.
-
NSCLC histology (squamous vs non‐squamous cell carcinoma).
-
Presence of molecular alterations (EGFR mutation, ALK and ROS1 rearrangements).
Sensitivity analysis
We will investigate the robustness of review findings by performing the following sensitivity analyses when appropriate.
-
Including only 'low risk of bias' outcomes, according to the summary assessment of risk of bias.
-
Including or not including results from studies with incomplete data, whether or not the data were imputed.
Methods for future updates
We will review the scope and methods of this review approximately yearly, or more frequently, if appropriate, in light of potential changes in the topic area or in evidence available for inclusion in the review (e.g. additional comparisons, interventions, or outcomes), or according to newly available review methods.
We will consider each year the necessity for the review to be living by assessing ongoing relevance of the question to decision‐makers and by determining whether uncertainty is ongoing in the evidence and whether further relevant research is likely.
